US7180818B2

High performance register file with bootstrapped storage supply and method of reading data therefrom

Summary by NHIP

Bootstrapped Multi-Port Register File

The storage array uses a row power supply driven above the main array supply voltage during access operations. Adjacent rows share specific row supply lines, with p-type field effect transistors acting as switches controlled by neighboring word lines.

Claim Score by NHIP

Read claim 33, the broadest

Abstract

A multi-port register file, integrated circuit (IC) chip including one or more multi-port register files and method of reading data from the multi-port register file. The supply to storage latches in multi-port register file is selectively bootstrapped above the supply voltage during accesses.

US7180818B2, drawing sheet 1 — Sheet 1 of 4

Term

Term ended

Expired 19 March 2025, 1.5 years ago.

39 claims: 4 independent, 35 dependent

1
A storage array comprising:an array of storage latches arranged in rows and columns;an array supply supplying power to said array;a row power supply at each row coupled to said array supply;a plurality of word lines, each of said word lines selecting a corresponding row of said storage latches and decoupling a corresponding said row power supply from said array supply;and a row selector selectively driving each of said plurality of word lines, drive from each selectively driven word line being coupled to said corresponding row power supply, said corresponding row power supply being coupled above said array supply.
13
An integrated circuit (IC) chip comprising:a multi-port storage array including a plurality of storage latches arranged in rows and columns;an array supply supplying power to said multi-port storage array;a plurality of word lines, each of said plurality of word lines selecting a row of said storage latches;a row supply at each row supplying power to said each row;at least one supply switch between said array supply and said row supply at said each row, power being supplied to said row supply through said at least one supply switch, a corresponding one of said plurality of word lines selectively opening each said at least one supply switch;and a row selector selectively driving each one of said plurality of word lines, said each one capacitively coupling a corresponding said row power supply above said array supply.
23
A CMOS integrated circuit (IC) chip including one or more multi-port register files, each multi-port register file comprising:a multi-port storage array including a plurality of storage latches arranged in rows and columns;an array supply supplying power to said multi-port storage array;a plurality of word lines, each of said plurality of word lines selecting a row of said storage latches;a plurality of row supply lines interleaved with and running parallel to said plurality of word lines, each of said plurality of word lines being capacitively coupled to at least one row supply line at a corresponding said row of latches, said corresponding row of latches receiving power from each coupled said at least one row supply line;at least one supply switch between said array supply and said each coupled at least one row supply line and, each supply switch being selectively opened by a corresponding one of said plurality of word lines;and a row selector selectively driving each one of said plurality of word lines, drive to said each one capacitively coupling each corresponding said at least one row supply line above said array supply.
33
Broadest claimClaim Score 78, broad(NHIP)A method of reading data from a storage may, said storage array arranged in rows and columns, said method comprising the steps of:a) selectively uncoupling a row supply line from an array supply;b) capacitively coupling voltage on said uncoupled row supply line above an array supply voltage;and c) sensing a data signal developing on one or more read data lines.

Independent claims4

22 paragraphs in 5 sections, as filed

FIELD OF THE INVENTION

The present invention is related to multi-port registers and more particularly improving multi-port register file performance.

BACKGROUND DESCRIPTION

Register files or, simply, registers are well known small, fast local storage arrays. A typical n by m register file includes storage latches in n rows and is m wide, e.g., a single byte, word or multi-word. Register files include, for example, first in first out (FIFO) or serial shift registers and first in last out (FILO) or push/pop registers. A FIFO may be a circulating shift register, for example, or a multi-port register with at least one input port and at least one output port. Additionally, typical such multi-port registers may be used for improving processor performance, e.g., in processor data queues or as pipeline registers.

In a state of the art pipeline structure, synchronous logic is segmented with a pipeline between segments or stages. So, in a pipeline processor, for example, a processor clock clocks pipeline registers distributed at strategic locations throughout the processor logic. Ideally, data latched in one pipeline stage propagates to, and arrives at, the next stage just as it is clocked into that next stage. So, pipeline registers act as boundaries between data units traversing the pipeline stages. Thus, for an N segment pipeline, N data units may be traversing the pipeline with one data unit in each segment. Also ideally, the logic delay through the N stages is N clock periods, i.e., the time each data unit spends in the pipeline is no more than necessary to propagate through the logic. So, ideal registers do not add path delay that detracts from overall performance.

In practice however, registers add to path delay, regardless of the register type (FIFO or FILO) or its use, e.g., whether as local storage or as a pipeline boundary. Consequently, for a pipeline circuit for example, the clock period limits the depth of the logic between pipeline registers to less than the clock cycle for any given clock frequency. Instead, the propagation delay between registers is offset or reduced by the register delay, where the register delay is the time through the registers, i.e., the time in and out of a register. So, the register delays reduce the time available for logic for each stage.

Further, the register delay is additive because it is encountered at each stage. For a pipeline circuit with 10 pipeline stages, for example, the 10 additional register delays may add one or more clock cycles to the time each data unit requires to traverse the pipeline, which is also known as the latency. Typically designers reduce the logic between stages with a corresponding increase in the overall number of stages to accommodate for these register delays. Each additional stage increases the circuit complexity without adding to the chip function; while it consumes valuable circuit area or real estate and so, reduces logic density. Further, each additional stage increases chip power, again without adding to the function and so, reduces chip efficiency. Of course, these problems dissipate as the register delays are reduced relative to other path logic.

Thus, there is a need for improved register performance.

SUMMARY OF THE INVENTION

It is a purpose of the invention to improve register performance;

It is another purpose of the invention to reduce register delays;

It is yet another purpose of the invention to reduce pipeline path latency.

The present invention relates to a multi-port register file, integrated circuit (IC) chip including one or more multi-port register files and method of reading data from the multi-port register file. The supply to storage latches in multi-port register file is selectively bootstrapped above the supply voltage during accesses, e.g., with a high K dielectric bootstrap capacitor.

BRIEF DESCRIPTION OF THE DRAWINGS

The foregoing and other objects, aspects and advantages will be better understood from the following detailed description of a preferred embodiment of the invention with reference to the drawings, in which:

<figref idref="DRAWINGS">FIG. 1A</figref> shows an example of a preferred two port storage register latch with a bootstrapped supply, such a may be included in a high performance N by M register file according to a preferred embodiment of the present invention;

<figref idref="DRAWINGS">FIG. 1B</figref> shows comparison of boosting cell supply voltage from three base supply voltages, 0.7V, 0.9V and 1.1V verses performance improvement for a typical register file cell;

<figref idref="DRAWINGS">FIG. 2</figref> shows an example of a cross section of a preferred embodiment high performance N by M register file of preferred embodiment cells.

DESCRIPTION OF PREFERRED EMBODIMENTS

Turning now to the drawings and, more particularly, <figref idref="DRAWINGS">FIG. 1</figref> shows an example of a preferred multi-port register cell 100 with a bootstrapped supply such as may be included in a high performance N by M multi-port register file according to a preferred embodiment of the present invention. The multi-port register cell 100 is a two port storage register latch in this example that, preferably, is in the insulated gate technology known as CMOS. The register cell 100 includes a pair of cross-coupled inverters 102, 104. A pair of write pass gate field effect transistors (FETs) 106, 108 are connected to the cross-coupled inverters 102, 104. A word-select line 110 is connected to the gates of write pass gate FETs 106, 108, which are connected between the cross-coupled inverters 102, 104 and a pair of complementary write lines, a write true (WRT) 112 and a write complement (WRC) 114. The cross-coupled inverters 102, 104, each include a p-type FET (PFET) 102p, 104p and an n-type FET (NFET) 102n, 104n and are connected between a word supply 116 and a supply return or register ground. Output 118 from one of the cross-coupled inverters (104 in the example) is connected to the gate of one of a pair of series connected NFETs 120, 122. The series connected NFETs 120, 122 are connected between ground (which need not be register ground) and a read data output line 124. A read-select line 126 is connected to the gate of the other of the series connected NFETs 120, 122. Thus, the cell contents 118 and the read-select line 126 are NANDed at read-data output line 124. A capacitor 128, preferably interline coupling capacitance, couples the read-select line 126 to the word supply 116. For additional or enhanced bootstrap capacitance, optionally, capacitor 128 may be a high K dielectric capacitor, e.g., with a high K dielectric material passivating the read-select line 126 to the word supply 116 wiring layer. Also, capacitor 128 may include an individually formed capacitor (e.g., an FET capacitor or conductive plates or alternating wiring layers) between each read-select line 126 and word supply 116.

Writing the register cell 100 begins with placing the intended data value on the pair of complementary bit write lines 112, 114, driving one high and the other low. Then, the word-select line 110 is driven high, which turns on the pass gate FETs 106, 108. Turning on the pass gate FETs 106, 108 couples the pair of complementary bit write lines 112, 114 to the cross-coupled inverters 102, 104. A single bit of data is transferred to the cross-coupled inverters 102, 104. Then, the word-select line 110 is returned low, which turns off the pass gate FETs 106, 108, latching the data in the cross-coupled inverters 102, 104.

Prior to reading data, however, the read-select line 126 is low and word supply 116 is at normal supply voltage, i.e., at Vdd. So, the full array supply voltage is applied to bootstrap capacitor 128, i.e., it is fully charged to Vdd. The read-data output line 124 is pre-charged high and may then be allowed to float. The stored data may be read out by driving the read-select line 126 high, which is NANDed with the contents of the cell, i.e., at the output 118 of inverter 104. So, with the read-select line 126 high, if output 118 is high, the read-data output line 124 is pulled low; or, otherwise, remains high. In addition however, the bootstrap capacitor 128 bootstraps the word supply 116 above Vdd, i.e., to Vdd+δ. The difference voltage is primarily a function of the ratio of bootstrap capacitance at bootstrap capacitor 128 and the apparent capacitance of the supply line 116, which includes direct and indirect (e.g., through on cross-coupled inverter PFETs 102P or 104P) cell capacitances. As noted hereinabove, in addition to line to line coupling capacitance, bootstrap capacitor 128 may include a space capacitor (e.g., an area capacitor of two plates on adjacent wiring layer or an FET capacitor) specifically added to enhance bootstrap capacitance. If cell contents internal node 118 is high, bootstrapping the supply voltage 116 facilitates switching the read-data output line 124, because Vdd+δ is passed to the gate of NFET 120, increasing the drive to switch the read-data output line 124.

<figref idref="DRAWINGS">FIG. 1B</figref> shows a comparison of boosting cell supply voltage from three base supply voltages, 0.7V, 0.9V and 1.1V at each of 130, 132 and 134 respectively, verses performance improvement for a typical register file cell, e.g., 100 in <figref idref="DRAWINGS">FIG. 1A</figref>. Thus, performance improvement may be realized by boosting cell supply in each of the examples 130, 132, 134 until an upper limit is reached, when the read performance improvement essentially plateaus, e.g., the added output transition time offsets the additional drive to a single FET (120) of the two NANDed devices 120, 122.

<figref idref="DRAWINGS">FIG. 2</figref> shows an example of a cross section of a preferred embodiment high performance N by M register file 140 of preferred embodiment cells, e.g., two port storage register latches 100 of <figref idref="DRAWINGS">FIG. 1A</figref>. In this example, 3 read word lines 126-0, 126-1, 126-2, of N words or rows are shown, with a single two port storage register latch 100 being shown in a single bit of one word, 126-1. A typical row driver 142-0, 142-1, 142-2 drives a corresponding read word line 126-0, 126-1, 126-2. Each of the N read word lines 126-0, 126-1, 126-2 are capacitively coupled to adjacent shared row supply lines 144, 146, 148, 150, which are shared in this example by cells connected to adjacent read word lines 126-0, 126-1, 126-2. Preferably, this capacitive coupling is from physical placement of the read word lines 126-0, 126-1, 126-2 interleaved with the shared row supply lines 144, 146, 148, 150, analogous to what is shown graphically in this <figref idref="DRAWINGS">FIG. 2</figref>. Thus, for example, the read word lines 126-0, 126-1, 126-2 and shared row supply lines 144, 146, 148, 150 may be physically located on a common chip layer at minimum pitch and organized substantially as shown. Each row supply line 144, 146, 148, 150 includes at least one switch 144-0, 146-0, 146-1, 148-1, 148-2, 150-2, 150-3 and etc., connected between the respective row supply line 144, 146, 148, 150 and an ungated supply line 152, e.g., register array or chip supply Vdd. In this example, except at boundary cells connected to boundary supply lines, e.g., 126-1, switches 144-0, 146-0, 146-1, 148-1, 148-2, 150-2, 150-3 are pairs of series connected PFETs in each end of the respective row supply line 144, 146, 148, 150. Each read word line 126-0, 126-1, 126-2 also gates off and on (opens and closes) the PFET switches 144-0, 146-0, 146-1, 148-1, 148-2, 150-2, 150-3 and etc. So, when a read word line, e.g., 126-1, is high, the corresponding row supply lines 146, 148 are decoupled from the array supply 152 because one of each respective pair is open, e.g., PFETs 146-1, 148-1 are off.

Data is written by placing the intended contents on complementary bit write pairs 112, 114 in <figref idref="DRAWINGS">FIG. 1A</figref> (i.e., driving one high and the other low) and driving a word select-line 110 high. Once data is written into the selected register location, the word-select line 1110 is dropped. A read is selected by driving low one input to the corresponding row driver, e.g., to 142-1. In response to the low on the input, the selected row driver 142-1 drives the corresponding read word line 126-1 high, which opens the row supply switches, i.e., turns off the connected pair PFETs 146-1, 148-1. With the switches 146-1, 148-1 open, the row supply lines 146, 148 are decoupled from the ungated supply 154. Simultaneously, charge across the bootstrap capacitance 128 couples the signal from the read word line 126-1 to the row supply lines 146, 148, bootstrapping them above Vdd. The higher cell contents 118 are NANDed with read word line 126-1. However, the higher bootstrap voltage is passed to 118, which causes read bit line 124 to switch faster than it would normally switch, i.e., unbootstrapped. Thereafter, the input to the row driver 142-1 is raised which drops the corresponding read word line 126-1, closing the switches 146-1, 148-1 in the row supply lines 146, 148 and reconnecting the row supply lines 146, 148 to the ungated supply 154. Thus, read performance is improved for each selected word, improving array performance.

Advantageously, bootstrapping the rows supply lines in the register, increases the read biases for improved storage register read time. Therefore, fewer pipeline stages are required for the same logic in a preferred embodiment pipeline. Thus, circuit and chip efficiency is improved and chip latency is reduced.

While the invention has been described in terms of preferred embodiments, those skilled in the art will recognize that the invention can be practiced with modification within the spirit and scope of the appended claims. It is intended that all such variations and modifications fall within the scope of the appended claims. Examples and drawings are, accordingly, to be regarded as illustrative rather than restrictive.

Contents5

Every citation, both ways

Citations
Document	Relation	Office	Cited during
US7259986B2	Cited by	United States of America	Search report
US7924633B2	Cited by	United States of America	Search report
US2010214859A1	Cited by	United States of America	Pre-grant
US2006215465A1	Cited by	United States of America	Pre-grant
US2004090818A1	Cites	United States of America	Search report
US6075741A	Cites	United States of America	Search report
US6549465B2	Cites	United States of America	Search report
US6597624B2	Cites	United States of America	Search report
US6671201B2	Cites	United States of America	Search report

4 members in 2 offices

This recordMember, by publication datePriority claim

Priority claims2

Priority claims
Document	Office	Kind	Date
99631104	United States of America	A	2004-11-22
US20040996311	–	–	–

Members4

Family members, oldest first
Document	Office	Kind
US2006109733A1	United States of America	A1	2006-05-25
CN1801389A	China	A	2006-07-12
US7180818B2This record	United States of America	B2	2007-02-20
CN1801389B	China	B	2012-02-22

29 transactions on the USPTO file

Allowed after 1 non-final rejection.

Non-final rejections: 1
Final rejections: 0
RCEs: 0
Appeals: 0

Over time

Point at a mark for the transaction

Transactions

Prosecution transactions, newest first
	Event	Code
2019-03-25	Expire PatentEXP.	EXP.
2018-10-08	Maintenance Fee Reminder MailedREM.	REM.
2007-02-20	Recordation of Patent Grant MailedPGM/	PGM/
2007-02-20	Patent Issue Date Used in PTA CalculationAllowedPTAC	PTAC
2007-01-31	Issue Notification MailedAllowedWPIR	WPIR
2007-01-17	Dispatch to FDCD1935	D1935
2007-01-17	Application Is Considered Ready for IssuePILS	PILS
2007-01-10	Issue Fee Payment VerifiedN084	N084
2007-01-10	Issue Fee Payment VerifiedN084	N084
2007-01-10	Issue Fee Payment ReceivedIFEE	IFEE
2006-10-13	Mail Notice of AllowanceAllowedMN/=.	MN/=.
2006-10-13	Notice of Allowance Data Verification CompletedAllowedN/=.	N/=.
2006-08-22	Date Forwarded to ExaminerFWDX	FWDX
2006-08-18	Response after Non-Final ActionA...	A...
2006-05-19	Mail Non-Final RejectionNon-final rejectionMCTNF	MCTNF
2006-05-15	Non-Final RejectionNon-final rejectionCTNF	CTNF
2006-02-16	Case Docketed to Examiner in GAUDOCK	DOCK
2006-01-10	Case Docketed to Examiner in GAUDOCK	DOCK
2005-02-01	IFW TSS Processing by Tech Center CompleteTSSCOMP	TSSCOMP
2005-02-01	Case Docketed to Examiner in GAUDOCK	DOCK
2005-01-03	Transfer Inquiry to GAUTI1050	TI1050
2004-12-22	Application Return from OIPEWROIPE	WROIPE
2004-12-22	Application Is Now CompleteCOMP	COMP
2004-12-22	Application Return TO OIPEROIPE	ROIPE
2004-12-22	Application Dispatched from OIPEOIPE	OIPE
2004-12-22	Application Is Now CompleteCOMP	COMP
2004-12-16	Cleared by OIPE CSRL194	L194
2004-12-15	IFW Scan & PACR Auto Security ReviewSCAN	SCAN
2004-11-22	Initial Exam Team nnIEXX	IEXX

15 legal events, as the office reported them to INPADOC

Over the term

Point at a mark for the event

Events

Legal events, newest first
	Event	Code
2021-05-12	AssignmentAS	AS
2021-05-12	AssignmentAS	AS
2020-11-20	AssignmentAS	AS
2019-04-16	Lapsed due to failure to pay maintenance feeLapsedFP	FP
2019-03-25	Lapse for failure to pay maintenance feesLapsedPATENT EXPIRED FOR FAILURE TO PAY MAINTENANCE FEES (ORIGINAL EVENT CODE: EXP.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITYLAPS	LAPS
2019-03-25	Information on status: patent discontinuationPATENT EXPIRED DUE TO NONPAYMENT OF MAINTENANCE FEES UNDER 37 CFR 1.362STCH	STCH
2018-11-29	AssignmentAS	AS
2018-10-08	Fee payment procedureMAINTENANCE FEE REMINDER MAILED (ORIGINAL EVENT CODE: REM.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITYFEPP	FEPP
2015-10-05	AssignmentAS	AS
2015-09-03	AssignmentAS	AS
2014-09-26	Fee paymentFPAY	FPAY
2014-09-26	Surcharge for late paymentSULP	SULP
2010-07-16	Fee paymentFPAY	FPAY
2006-12-01	Fee payment procedurePAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITYFEPP	FEPP
2004-12-14	AssignmentAS	AS